2 ms·
Take a look at https://replicate.com/blog/fine-tune-llama-to-speak-like-homer-simpson https://replicate.com/blog/fine-tune-llama-to-speak-like-hom... (and the r
by joshka 3y ago
Take a look at https://replicate.com/blog/fine-tune-llama-to-speak-like-homer-simpson https://replicate.com/blog/fine-tune-llama-to-speak-like-hom... (and the related blog posts) for a good example of this.
- Animats 3y agoWell, OK. They put in 61,000 lines, the entire first 12 seasons of the Simpsons's. Lines such as {'previous': 'Marge Simpson: Ooo, careful, Homer.', 'character': 'Homer Simpson', 'line': "There's no time to be careful."} and then used it to generate more Simpsons dialogue. With a data set so well matched to the goal, the algorithm barely matters. Remember that recent paper, "Copy is all you need."?[1] This is the ideal case for that approach. I'm thinking more in terms of loading in the detailed product descriptions and maybe manuals from a catalog, then letting users ask questions about how to do things with the products. [1] https://news.ycombinator.com/item?id=36758233 https://news.ycombinator.com/item?id=36758233